Predicting patient knowledge of cardiac risk factors: A comparison of two approaches

Abstract

The University of Ottawa Heart Institute conducted a survey of patients to understand the level of knowledge of cardiac risk factors and to identify any subgroups of patients that could benefit from specially designed educational programs. This thesis compares two approaches to the analysis of this multidimensional dataset. Both techniques looked at the 10 modifiable risk factors and a number of predictor variables (age, gender, education, and smoking status). Logistic regression was hampered by low sample size, sparse data, and the high probability responses of many of the binary knowledge variables. Only one risk factor was successfully explained by any of the predictor variables. Correspondence analysis demonstrated that those who are unaware of smoking as a risk factor are not current smokers; knowledge of low fibre diet is related to education but not to gender; females are more aware of high salt diet and stress as risk factors, and are more likely to have never smoked; smokers tend to have lower education and be unaware of the risk of a low fibre diet.